Here’s to resume the topic of sibilant+stop clusters in loanwords into Finnic, previously treated in “Extending a substitution pattern”. This time I’m focusing on clusters involving the bilabial stop /p/.
For some reason, clusters of the type /Sp/ were originally alien to the Uralic languages. The comparative material contains an abundance of sibilant + *k clusters (e.g. *puskə- ‘to push’, *mośkə- ‘to wash’, *päškə ‘nut’); the evidence for sibilant + *t is however scarcer (examples like *wasta ‘against’ or *toštə- ‘to dare’ exist, but these are limited to the more Western branches ); and no old cases of sibilant + *p cover more than one Uralic subgroup, if even that. This is a curious unexplained situation.
This post will investigate not Proto-Uralic though, but newer a specific issue: a particularly “sušpiciously” missing cluster of this type. While /sp/ is by now found in modern Finnish (if still rare), the related cluster ˣ/hp/ for some reason remains impossible. A major reason probably is that while /sp/ can originate in loans from adjacent Indo-European languages (as even a marginal glance at words containing this cluster shows: piispa ‘bishop’, raspi ‘rasp’, colloquial respa ‘reception attendant’), this kind of a direct origin is not possible for /hp/.
However, this is unsatisfying as the entire explanation. Around the Early to Middle Proto-Finnic period, a relevant non-trivial regular substitution pattern in loanwords was Proto-Germanic *s being adapted as Finnic *š.  If any words with word-medial *sp had been loaned from Germanic around this time, they should have yilded Late Proto-Finnic *hp, just like PGmc *askōn yields dialectal Finnish ahku ‘ashes’, or PGmc *wiskō- ‘whisk’ yields Finnic *vihko ‘bundle or whisk for washing’. Yet there are no examples around. 
This might not be an accident: I have gotten together some reasons to believe a cluster *šp did exist in Middle Proto-Finnic (MPF), and that it has evolved into something else than the expected ˣ/hp/.
The first word to consider here is ‘aspen’: *haapa, found all across the Finnic languages (though South Estonian has unexpected -v-: haav). This can be readily compared with the PGmc name of the tree, *aspōn. If this had been loaned into MPF, we would predict it to have been rendered as *ašpa, and then Late Proto-Finnic *ahpa. I propose a metathesis would have occurred additionally, yielding a /hVp/ structure rather than a cluster /hp/.
I am not committed to a specific ordering for this metathesis and the shift *š → *h. There is some reason to suspect that the metathesis would have been early: namely, a very similar word for ‘aspen’ is also found in Mari, as West /šapki/ ~ East /šopke/ (this mostly adds up to a regular Proto-Mari form *šapkɜ, but I don’t know what to do with the final vowel that varies as i ~ e). A proto-form *šapa has been reconstructed in sources such as the UEW; but instead of inheritance, this might rather be an early loan to Mari from a para-Finnic offshoot in which *š was retained. Such an early eastward expansion of MPF has been previously posited to explain loanwords of Germanic or Baltic origin, or words showing characteristically Finnic features, turning up in the nearby Permic languages. So why not also Mari?
As for the long vowel *aa in Finnic, I could suggest that this is an indication of the formerly syllable-final, mora-carrying consonant; but there are also some hints of irregular lengthening *a → aa after /h/, in at least haara ~ hara ‘tree branch’, and haalea ~ halea ‘lukewarm’.  The same might also apply here. Note that this assumption would be also required by any etymology according to which the Finnic and Mari words are related by descent, and that it therefore does not count as a weakness of my loan etymology in particular.
A third Uralic word for ‘aspen’ that might belong here is Samic *supē, suggesting earlier *sVpa or *šVpa. I’m not aware of any good explanations for the *u though, so I’ll leave this one aside for now.
Once again, a single example could always be considered simply an accidental similarity though. (My rule of thumb: one example is no example, two examples is a pattern, three examples is a sound law.) So is there any other potential evidence for this supposed metathetical dissolution of *hp (or *šp)?
The answer is yes, provided that we look a little farther out! I believe this idea also provides an etymology for Finnic *hëpo ‘horse’. This word, which has no accepted farther-off cognates, has for long been considered to show intriguing similarities to the Indo-European ‘horse’ words, cf. e.g. Greek hippos. But no plausible loan etymology has been yet put together. Direct comparison with PIE *h₁eḱwos might suggest correspondences *h₁ : *h, and *ḱw : *p. Yet these are probably illusory. 
If *hëpo rather goes back to Early Proto-Finnic *ešpo (or perhaps *ešpa-w at this time level), it appears to be possible to explain this as having been loaned from some Indo-Iranian source. The PII form *aćva perhaps does not look like an immediate match, though. One possibility is that the donor was some early northwestward Iranian(ish) offshoot, in which *v → *p had occurred following a voiceless consonant (much like in Avestan aspa). But *p could also have been directly substituted for *v for phonotactic reasons: **čw or **šw were not possible consonant clusters in Proto-Uralic or its immediate descendants. 
At a sufficiently old date, one might expect the substitution *ć → *ć ~ *ś; but several parallels, including Permic *už ‘stallion’ from this very same root, confirm that whatever language(s) exactly was (were) responsible for the old II loans in Uralic, a postalveolar-ish value for PII *ć occurred in it (them). And although the usual notation I’ve seen for PII is *ć (*ĉ) from PIE *ḱ (*k̂), and *č from secondary palatalization of PIE *k⁽ʷ⁾, it was probably the newer consonant, i.e. the latter, that was more strongly palatal. The former would have to have already shifted out of the way for the two to not have merged immediately. That is, we would seem to have two notations here that are the inverse of each other: PII *ć, PU *č = postalveolar [tʃ], versus PII *č, PU *ć = (alveolo-)palatal [tɕ ~ c]. The *š [ʃ] of my EPF reconstruction would then be a natural intermediate between *[tʃ], and the /s/ that is attested all across Iranian. (As for PII *š from RUKI, this was probably retroflex [ʂ], and does not necessarily get in the way.) It could in principle of course also be a Finnic-internal development from older *-čp-; but since *-čk- (→ MPF *tk) remains distinct from *-šk- (→ MPF *šk → LPF *hk), maybe not.
The vocalism might seem to point to an even older date. Traditionally, any old Satemic loans into Uralic that appear to show a front vowel have been explained as loans from pre-Indo-Iranian, before the shift *e → *a. There are some problems in this approach though. For starters, the distribution of these *e-loans does not suggest a particularly early date. Most that I’ve seen are found in Finnic + maybe one other branch. This is in contrast to several back-vocalic loans such as *śëta ‘100’, *aja- ‘to drive’ that cover just about the entire Uralic family. There are even examples of a competing *e-loan (e.g. PIE *jewa → EPF *jewä → LPF *jüwä ‘grain’) and a slightly more widespread *a/*o-loan (e.g. PII *java → dialectal PU *jawə or *jowɜ → Mordvinic *juv, Komi /jɨkɨ/, Udmurt /ju/ ‘id.’), which at least to me suggest that these were acquired in parallel, not at different times.
Another issue is that some of these front-vocalic loans actually show distinctly Iranian-style post-PII developments. As covered above, I do not suppose that *č-like reflexes (including the current example) must be counted among them — but examples with regular alveolar *s also exist. This has led to J. Koivulehto to propose that the change *e → *a was actually very late, perhaps even later than Proto-Iranian! I can only assume this would leave II-ists puzzled.
A better solution may be available. Again, there are also well-known loanwords in which PII *a corresponds to Uralic *ë or *o, and the received explanation for most of these is different: PII *a, as the only non-close vowel in the language, would have been sufficiently variable in pronunciation that it could have been substituted by several different vowels on the Uralic side (especially when PU *a was probably firmly stashed in the [ɑ ~ ɒ] corner of vowelspace). I suspect that it is this phenomenon that is behind at least a part of the *e-loans as well. In particular, *aćva has *a occurring before a rather palatal consonant, which seems likely to induce a more fronted value for the vowel.
As I’ve mentioned before in the comments section, I actually have one new *e-loan proposal forthcoming with a fairly wide distribution. Yet this one would actually only work against idea that a Uralic *e must indicate a PIE *e-grade — since this time no *e-grade is attested in the PIE original. A cookie for any reader who manages to realize what word I’m talking about before I release the details on this etymology…
In summary: it appears that two Finnic words that remain without a proper etymology (*haapa ‘aspen’, *hëpo ‘horse’) can be explained as IE loanwords (from Germanic *aspōn, II *aćva resp.), by assuming a metathesis of *h (or earlier *š) from before *p to the word-initial position. While this is not a large pool of evidence, some minor supporting evidence comes from that this sound change also provides a (partial ) explanation for the absense of the cluster /hp/ from the modern-day Finnic languages.
 Actually I wonder if *wasta, found in West Uralic + Mari, could be connected with the PU root *wasa ‘left’. The Finnish reflex vasen of the latter is of the wrong stem type compared to Samoyedic *wåtå. This could be explained if these were instead parallel derivatives from a root *wasə-, such as pre-Finnic *wasə-m(ə), vs. pre-Smy *was-ta. Since *s goes to *t in Samoyedic, and no geminates are found in PSmy either, *wåtå would still be the expected reflex. This could then be identified with *wasta ‘against’, i.e. meaning ‘the hand opposite to the right’? This opens too many questions of morphology and semantics for me to pursue this idea further right now however.
 This probably incidates that, much like modern Finnish /s/, Proto-Germanic *s was pronounced as a slightly retracted but not quite fully postalveolar sibilant: [s̱]. This is common in languages where the “plain” sibilant does not contrast with an explicitly postalveolar one. *s before consonants yielding Modern Standard German sch [ʃ] may be a testament to this. This phenomenon also accounts for many “unexpected” shibilants in loanwords across Europe; IIRC it is, for example, why Middle French pousser *[pus̱e(r?)] was adapted as English push [pʊʃ]. I’ve got a full paper on the topic somewhere in my files, feel free to ask if you’d like me to dig up the details.
 I actually do not recall examples with *st → *ht existing either. This PGmc cluster was probably pronounced something like [s̪t̪], and so here the substitution as Finnic *st (or, word-initially, *s) persisted. — It may be relevant to note that in Modern Finnish, vihko has evolved to mainly mean ‘notebook’ (= ‘bundle of pages’). Meanings such as kukkavihko ‘flower bouquet’ were still in regular use some decades ago.
 Seemingly also haarniska ‘armor’ from Old Swedish harnisk (and further related to English harness), but I think this is rather an example of regular lengthening of *a before /rn/ (also appearing in e.g. saarna ‘sermon’, saarni ‘ash tree’, kaarna ‘bark’, vaarna ‘peg, stake’; there is also an irregular example before /rm/, paarma ‘gadfly’); and haahka ~ hahka ‘eider’, which seems best explained by lengthening before coda /h/+stop, a change widespread in Karelian and Savonian but only sporadically adopted in standard Finnish (some other examples include hiihtää ‘to ski’, paahtaa ‘to roast’, vaahtera ‘maple’, vaahto ‘foam’).
 There is no known relationship between Finnic *h-, and PIE *H-. In those proposed loanwords where PIE laryngeals would be reflected as consonants in Finnic, the sole word-initial correspondence is PIE *H- : PF *k-. Word-medially, a correspondence PIE *-HC- : MPF *-šC- → LPF *-hC- has been proposed, but I am somewhat skeptical on this interpretation. More later on this perhaps.
 In more general, no combinations of voiceless consonant + glide were. MPF did have the clusters *-tv-, *-sv-, *-tj-, *-cj-, *-kj-, *-pj-, and I suspect perhaps *-kv-… but only the first of these is inherited, arising from PU *-d₂w-, where *d₂ might have been a liquid rather than an obstruent. Also *-sv- appears in at least one fairly old loanword: *kasva- ‘to grow’, with a Mordvinic cognate *kasə-, but here a loan etymology from PIE *h₂awks- requires a metathesis from EPF *-ws-. All the others occur solely in newer loanwords from Germanic and Baltic.
 It is merely a happy accident that both of the etymologies proposed here start from a vowel-initial loan original, required for my suggested metathesis to be possible even in principle! It seems difficult to derive any prediction from these on what a hypothetical *CVšpV root in EPF would have yielded in the present day.
What about the -sv- in Finnic usva “mist”, Sami osve “wet, clinging snow”? (If we assume that these are cognates to begin with, that is.)
They seem likely to be related, but the distribution does not suggest common inheritance: these are attested only in Northern and Inari Sami on one hand, Finnish and Karelian on the other. I’d guess they’re from the Samic substrate in Northern Finnic.
There’s a Finnish dialectal variant usma, but this could be 2ndary, due to the influence of sumu “fog”. A Southern Karelian variant udžve could also be simply expressive variation, but it also could point to the Proto-Samic form being *ocvē rather than *osvē (preconsonantal *c is merged into /s/ widely across Samic, including NS and IS).
Pursuing this idea further actually suggests that a cognate might be Finnic *utu “mist, haze” instead. Connecting the 2nd syllable *-vē (← *-wa?) in the former with *-u (← *-əw?) in the latter is difficult; but aside from that, a root beginning *uč- would regularly yield Finnic *ut- ~ Samic *oc-. South Estonian udsu also seems to point to previous *č, although normally *-č- merged with *-t- in the language, as in the rest of Finnic. (Speculation: this could indicate a former syllable-final position, since *čk mostly regularly yields SE /tsk/, and that *učvu rather than *uču should be reconstructed for MPF?!)
So, do you think the Sanskrit ś is secondarily fronted (perhaps to maximize contrast with the retroflex)? – Of course, this has little bearing on your argument, because any loans into Uralic more likely came from Iranian than from Proto-Indo-Iranian.
IPA [c], BTW, is a plosive, not an affricate. The Sanskrit shift from *č to c has been ascribed to Dravidian influence, IIRC.
Well, Avestan has s while Old Persian has what is transcribed θ… and the Nuristani languages, closer to Indic than to Iranian, all have [ts] if I read Appendix 2a of this paper correctly. (Complete list of consonant correspondences between PIE, PII, Vedic Sanskrit, Avestan, Old Persian, Proto-Nuristani, and all 5 extant Nuristani languages.)
…Well, yes, it would.
Not necessary, it’s all here on Wikipedia! :-) (Make sure to scroll down to the “Occurrence” section.) Incidentally, I found this article via the one on Proto-Semitic, which probably used this sound for what is traditionally transcribed *š (while *s was [tsʰ]).
Yes, I’d suspect so. This would parallel nicely the merger of both voiced affricates to a palatal rather than a postalveolar.
I’m aware, yes.
Also thanks for linking the paper on Nuristani’s position. Makes for good comparision with a site that was recently brought up at the ZBB: Nuristan.info, which among other things treats the same topic in some respects in more detail.
“Where is Nuristân, and Who Cares?” :-D